Feeding Market Signals into an AI is not calculating a profitable price
A market signal describes what competitors are doing. A profitable decision tells you what you need to do, and how it will affect sales, margins, and inventory. The two are not part of the same tool.
The decision-making process relies on five key components that external data does not provide: sales forecasts, price elasticity, cannibalization, business rules, and impact simulation. A signal infrastructure feeds these components; it does not replace them.
A company specializing in scraping and matching now bills itself as “the intelligence layer for agent-based retail.” The message is simple: a single data infrastructure for pricing, product content, buying agents, and retail media.
The promise is appealing. However, it blurs the line between two distinct tasks: feeding market signals into an AI and calculating a profitable pricing decision.
This article is intended for pricing, sales, and data teams that are selecting a tool. It explains what a market signal provides, what it does not reveal, and what a pricing engine must calculate in addition.
Market signals and profitable decisions: two different fields
Feeding market signals into an AI involves collecting, reconciling, and updating external observations: competitor prices, availability, promotions, and product listing content. The result is data.
Calculating a profitable decision involves transforming this data—and, most importantly, your own sales data—into a recommended price.
The impact of this price on volume, margin, and inventory is calculated before it is applied. The result is an arbitrage.
The first activity involves data collection and product matching. The second involves data modeling. They complement each other, but the second does not automatically follow from the first.
A competitor's price recorded every ten seconds is still just an observation. It tells you neither how much you will sell at that price, nor what it will cost you in terms of your margin, nor which related product will suffer as a result.
The data layer aims to become the infrastructure for agent-based retail
The move is real. Intelligence Node, a long-standing player in rate monitoring and matching, was acquired by Interpublic in late 2024 and subsequently integrated into Omnicom following the merger of the two groups, which was finalized on November 26, 2025.
The company now describes itself as “The Intelligence Layer Powering Agentic Retail.”
Its offerings include pricing, product data, product assortment, content, and AI-powered catalog preparation. It draws on the sales and data capabilities of a global communications group.
Why it's appealing: Purchasing agents come in and review product listings, prices, and inventory data that are organized in a structured format. A brand or a major retailer may want a single infrastructure to feed all of them.
Traffic from AI sources to U.S. retail websites in the first quarter of 2026 compared to the first quarter of 2025. In March 2026, this traffic converted 42% better than traditional traffic.
(Adobe Digital Insights, April 2026, more than 1,000 billion visits analyzed)
According to Salesforce, AI and agents influenced 20% of global online sales during the 2025 holiday season, amounting to $262 billion. The data layer is becoming strategic. We analyzed this in our article on agent-based pricing.
But let's take a closer look at the promise.
On its page dedicated to dynamic pricing, Intelligence Node writes that pricing engines “respond to competitor price movements that they can observe.”
And the company describes itself as “the high-velocity competitive data layer behind yours” (translation).
In other words, the engine remains the client's. The data layer provides the inputs for its rules and scenario tests. It does not calculate the decision, nor does it claim to do so.
What a Market Signal Doesn't Tell You
A market signal answers one question: What are others doing? A pricing decision raises five more, and each one requires a model built on your data.
Of the executives surveyed, most use AI in pricing primarily to gather market intelligence. Only 28% use it to optimize prices in real time.
(Simon-Kucher, Global Pricing Study 2025, 2,200 executives, 28 countries)
This discrepancy is no coincidence. Collecting data is simpler than modeling it. The recency of data is a real advantage, as we have shown with regard to up-to-date competitive data, but it does not replace any of the subsequent building blocks.
Sales Forecasting: How Much Will You Sell?
Before adjusting a price, you need to know how many units the product would have sold without that change. Sales forecasts take into account seasonality, events, weather, and the sales history of each SKU at each retail location.
Without it, no effect of a price can be measured. We're comparing a number to nothing.
Price elasticity: How does your demand respond?
Price elasticity measures the change in volume in response to a change in price. It varies from product to product, from distribution channel to distribution channel, and sometimes from season to season.
A competitor's price says nothing about your price elasticity. Matching a rival's price reduces your margin when demand is inelastic, and isn't enough when it's elastic.
Cannibalization: Who Pays the Price for the Decline?
Lowering the price of a product shifts sales from neighboring shelf items, from national brands to private labels, and from large-size to small-size items. The opposite effect also exists; this is known as the halo effect.
These cross-elasticities cannot be derived from any external data. They are calculated from your sales history, category by category.
Business Rules: What Is Prohibited?
Price tiers across product lines, consistency across formats, psychological pricing, minimum margins, and the regulatory framework: a recommendation that ignores these constraints is not feasible.
Rules provide a framework for calculations; they do not replace them. An engine built solely on rules ends up with a "rule debt."
Impact Simulation: What Happens If...?
The final building block ties the other four together. Before applying a scenario, the simulation quantifies its impact on revenue, margin, price image, and inventory, item by item.
That is what distinguishes a recommendation from an opinion. A data layer can provide the competing scenarios to be tested. The test itself is performed elsewhere.
Market Signal or Decision Driver: A Comparison
The table below summarizes what each layer can answer. The point is not to pit two tools against each other, but to determine which one answers which question.
| Question | Market Signal Layer | Pricing Decision Engine |
|---|---|---|
| What are the competitors doing? | Yes, that's what it's all about: prices, inventory, promotions, and content | Yes, as input for the calculation |
| How many will you sell at that price? | No | Sales Forecast by SKU and Store Location |
| How does your application respond? | No | Intrinsic elasticity, measured based on your sales |
| Which related product is affected? | No | Cross-elasticities, cannibalization, halo effect |
| Is the recommendation applicable? | Background Information on Your Period | Business rules integrated into the calculation |
| How will this affect the margin and inventory? | Scenarios to Test | Cost Estimate Simulation Before Implementation |
The first two columns overlap in a single row: knowledge of the competition. Everything else is handled by the decision engine, and that’s where the margin is determined.
Why Scientific Depth Can't Be Bought with Data
It's tempting to believe that decisions are driven by data. With enough signals—and fresh enough ones—the right price would emerge on its own. Experience shows the opposite: the bottleneck isn't data collection; it's the model.
Modeling elasticity across thousands of SKUs—with promotions that skew historical data and stockouts that mask demand—is a data science task. It requires years of hands-on experience in the field.
Planned price increases are actually being implemented, a rate that has fallen by 5 points over two years. The gap between pricing goals and execution is widening.
(Simon-Kucher, Global Pricing Study 2025)
At Booper, this depth of expertise has been at the heart of our business since 2014. The team includes four PhDs in AI and data science, and sales forecasts achieve 95% accuracy across the areas we manage.
A national food retailer with more than 1,700 stores has thus transitioned from reactive pricing to predictive pricing.
Simple and cross-pricing, cannibalization, pre-execution simulation, validation rules, and workflows are all combined in a single pricing solution.
Market signals are fed into the system as raw data through price monitoring and product matching. They never result in recommendations on their own. The entire mechanism is described in “How an AI Pricing Engine Works.”
How this affects your choice of tool
A single data infrastructure for pricing, content, and retail media makes sense, especially for a brand that needs to provide data to buying agents. It addresses the issue of visibility.
It doesn't address the issue of profitability. Ask to see how the tool forecasts your sales, measures your price elasticity, and quantifies the impact of a recommendation before it's implemented. If the answer refers to "your engine," you have your answer.
Good architecture is often a combination of both: broad, up-to-date data collection and a decision engine that uses that data to drive profit margins. One without the other leaves half the work undone. That is the whole point ofprice optimization.
Would you like to see the impact of a price before applying it?
Our experts will use your own data to show you the forecasting, elasticity, and impact simulation behind each recommendation.
Let's discuss your decision-making processFAQ: Market Signals and Pricing Decisions
A price monitoring tool collects and cross-references external data—such as competitors’ prices, inventory levels, and promotions—and links them to your product listings through product matching. It answers the question, “What are others doing?”
Pricing optimization software uses these observations and your own sales data to calculate a recommended price. It forecasts volume, measures price elasticity, takes into account cannibalization and business rules, and then quantifies the impact before implementation.
The first produces data; the second produces a decision.
No. A feed of competitors' prices—even if updated every few seconds—says nothing about your demand. It indicates where the market stands, not what you'll sell at a given price or what that will cost you in terms of your margin.
Dynamic pricing that simply follows competitors amounts to delegating your pricing policy to them. To make a decision, you need sales forecasts, elasticity measured based on your own data, and a simulation that quantifies each scenario before it is implemented.
A market signal is an external observation that describes a product’s competitive environment: a competitor’s listed price, availability, current promotions, ranking on a comparison site or marketplace, and the content of the product listing.
These signals are collected through price surveys and web scraping, then linked to your product listings through product matching. They serve as input for the pricing engine.
They do not constitute a decision, because they say nothing about how your customers will react to a price change.
Because an effect can only be measured relative to a baseline. To determine the impact of a price change, you must first estimate how much of the product would have been sold without that change, at each retail location and over the period in question.
Sales forecasting provides this benchmark. It takes into account seasonality, events, weather, and the historical data for each SKU. Without it, a pricing recommendation has no quantifiable impact and therefore no economic justification.
Cannibalization refers to the shift in sales from one product to another within your own product lineup when you change a price: from a national brand to a private label, from a large size to a small size, or from one product line to a related one.
This phenomenon occurs within your catalog. It cannot be detected through external observation. It is measured using cross-elasticities calculated based on your sales history, category by category.
He often reverses a decision that initially seemed profitable, on a case-by-case basis.
Business rules define the constraints that every recommendation must adhere to: pricing tiers across product lines, consistency across formats and channels, psychological pricing, minimum margins, and industry-specific regulatory requirements.
They provide a framework for the AI's calculations and ensure that its recommendations can be implemented without manual adjustments. They do not replace the AI. An engine consisting solely of stacked rules quickly becomes unreadable, and no one knows why a price is what it is.
The impact simulation quantifies, prior to implementation, the effect of a pricing scenario on volume, revenue, margin, price perception, and inventory—item by item and store by store.
It combines sales forecasting, own elasticity, cross-elasticities, and business rules. It is this process that transforms a recommendation into a well-reasoned decision and allows for the comparison of multiple scenarios before selecting one.
It makes it more visible. Purchasing agents review structured product listings, prices, and inventory levels, and clean, up-to-date data has become essential for appearing in their search results.
A single infrastructure for pricing, content, and retail media therefore makes sense.
But visibility doesn't equal profitability. When dealing with an agent who compares offers at lightning speed, knowing how low to go without eroding the margin matters more than ever. That answer comes from the decision engine, not from the stream of signals.
When choosing sales forecasting software, there are five key points to check before conducting any tests: demand hidden by stockouts, the forecast granularity, an explanation of each figure, the relationship with prices and promotions, and an honest assessment of accuracy.
A demonstration doesn't prove anything on these points. Only a test using your own data, compared to a simple baseline, will show whether the forecast truly improves your decisions.
A pricing proof of concept (POC) does not require perfect data, but rather data that is appropriate for what it needs to demonstrate, with known limitations. The product catalog, sales data, prices and promotions, purchase costs, and inventory levels form the foundation; the granularity and depth of this data depend on the test’s objective.
Before calculating a profit, the recalculated revenue, volumes, and margin must match those in management accounting. A known and documented error does not skew the comparison; a hidden error does.
The model is the easy part of a pricing tool. What takes years is the product catalog, product matching with competitors, rules, explainability, and compliance—some of which have expiration dates. Across all tools, 71% of in-house developments end up being abandoned, according to Exclaimer (2025).
The hidden cost is that of continuity: an internal tool relies on two or three people, and it is its success that creates dependency. The build process remains meaningful within a narrow and stable scope, as a complement to a platform. The deciding factor is not the ability to build, but the ability to maintain it for ten years.